This project focuses on an emerging challenge in computation: to extend programmatic control over matter and phenomenon at the nanoscale. Nanosystems making use of DNA-based reactions are a promising technique to achieve this since they are feasible to design, simulate and test experimentally. DNA computation systems of increasing complexity have been demonstrated over the past two decades. Most of these systems involve multiple strands of DNA that interact with each other via diffusion based hybridization chemistry. While this paradigm has many advantages and merits, there are fundamental limits to diffusion based DNA hybridization computations, particularly due the increased time for larger-scales of complexity. This work seeks to study an alternate paradigm of DNA hybridization-based computations that operate locally on a substrate. Locality allows reactions to proceed at higher speed due to increased local concentration of reacting species - this localization could potentially speed up DNA hybridization-based computations by an order of magnitude. Also, since each of the local reaction pathways do not interfere with each other, it is also possible to simultaneously execute multiple pathways in parallel. This also allows one to reuse DNA sequences in spatially separated regions that increase the modularity and scalability of the reactions.
Intellectual Merit: The research work spans both theory and experimental techniques, and includes development of biophysical mathematical models, design software, computational simulations, small-scale experimental demonstrations. In particular, the work will develop biophysical models of localized hybridization, which will be simulated, and also verified via simple kinetic experiments. The experiments provide crucial data about the rate constants involved in the hybridization chemistry of localized molecules. The simulation model will be further refined based on the experimental data,. A major challenge addressed as a center-piece of this effort is leaks: the unintended reactions that cause the nanosystem to significantly deviate from its programmed trajectory that might occur in localized hybridization systems. Multiple leak models will be tested in the lab via simple experiments. Continuing an on-going collaboration with Dr. Andrew Phillips (Microsoft Research Cambridge), funded internally by Microsoft, this work will also create software systems that will simulate localized hybridization networks. The simulation software development will be tightly coupled to the experimental progress by constantly refining the simulation models and parameters based on experimental data. Finally, this work will experimentally implement a series of small to moderate scale localized hybridization systems to demonstrate the feasibility and the potential of localized hybridization reactions. The work will also investigate the broader issues of the use of locality to speed-up other related molecular-scale computation processes, including reactions that make use of enzymes, or other protein-based reactions, in addition to DNA hybridization reactions.
Broader Impact: There is substantial multidisciplinary impact to nanoscience, biochemistry and chemistry, which will profit from the introduction of key methodologies derived from mainstream computer science, such as mathematical modeling, software engineering, algorithms and modular design methodologies. Educational impact includes cross-disciplinary training of four PhD students, carefully supervised mentoring and summer internships for undergraduates.
